https://www.promptquorum.com/api/mcp
स्थानीय AI जानकारी तक पहुँच प्राप्त करें
दस्तावेज़
PromptQuorum Speaks MCP
PromptQuorum runs a public, read-only MCP server, so any MCP-compatible AI assistant — Claude, or any other client speaking the open Model Context Protocol — can search the directory and guides, recommend apps for your hardware, and compare tools directly inside your own chat, without you opening a browser tab.
Connect your AI assistant → https://www.promptquorum.com/api/mcp
How it works
Connect once, then ask in your own words. Your assistant picks the right tool, calls it, and answers with links back to the source pages. You do not need to know tool names — they are listed here so you can see what is possible.
What it can do
Find apps
- find_best_apps Find the best apps for you Turns a plain-language goal plus your operating system and memory into a ranked shortlist, with reasons and honest trade-offs. Ask: I have a 24 GB Mac and want to chat with my PDFs without using the terminal.
- search_apps Recommend apps Recommend local-AI apps from the directory that match a stated goal and hardware. Ask: Free Windows apps for image generation that run on 8 GB of VRAM.
- recommend_stack Recommend an app stack Suggests a ready-made combination of apps for a goal, such as Ollama + Open WebUI, together with its hardware floor. Ask: Which apps should I combine for a private coding assistant with 32 GB of RAM?
- find_local_alternative Find a local alternative Maps a cloud AI product such as Midjourney or ElevenLabs to local alternatives, and states the gaps honestly. Ask: What is a local alternative to ElevenLabs?
- get_app_alternatives Find app alternatives Finds similar apps when one is too hard, too costly or does not fit your machine. Ask: ComfyUI is too complicated for me — what is easier?
- get_app_details Get directory entry Fetch license, hardware requirements, platforms, price and URL for one tool. Ask: What are the license and hardware needs of Ollama?
- list_categories List categories List directory categories, use cases and operating systems. Ask: Which app categories exist, and how many apps are in each?
Decide and check
- compare_apps Compare apps Compare local-AI apps side by side on license, hardware, platforms and price. Ask: Compare Ollama, LM Studio and Jan for a beginner.
- check_hardware_compatibility Check hardware compatibility Says whether an app — and optionally a specific model size — fits your platform and memory, and what to try instead if it does not. Ask: Will ComfyUI run on my Windows laptop with 8 GB of VRAM?
- estimate_vram Estimate VRAM Estimates the memory a model needs from its size, quantization and context length, and whether it fits what you have. Ask: How much memory does a 32B model need at Q4 with an 8K context?
- explain_license Explain a license Explain the core rule of a software license (AGPL-3.0, Apache 2.0, MIT, …) in plain language. Ask: Can I use an AGPL-3.0 app inside my company's product?
Read and learn
- search_promptquorum Search PromptQuorum Search local-LLM guides, prompt-engineering articles, and the software directory. Ask: How do I run Llama on a Mac?
- get_article Get PromptQuorum article Fetch the full content of one article by cluster and slug. Ask: Open the full guide you just found.
- Collects the review, comparison guide, hands-on test and setup articles for an app or topic. Ask: Show me the review and setup guides for Open WebUI.
- get_hands_on_test Hands-on test findings Returns PromptQuorum's own hands-on findings for an app, with evidence labels — or says plainly that it was not tested. Ask: Did PromptQuorum actually test BoBe, and what went wrong?
- get_latest What's new Lists recently added apps, updated articles and published hands-on tests, newest first. Ask: What was added to the directory recently?
- list_clusters List content clusters List every content section on the site with a short description. Ask: What topics does PromptQuorum cover?
Start and your profile
- start Start here Opens the guided menu: a short privacy question first, then what this connection can do and a numbered list of tasks. Nothing is saved until you say yes. Ask: Start
- Records your yes or no. Yes saves a small pseudonymous profile (language and consent record) under a random code; no saves nothing and every directory tool keeps working. Ask: Yes, save my language and setup.
- my_profile Show my profile Shows exactly what is stored for your profile code: consent date, consent text version, language, timestamps and expiry. Ask: What do you have stored about me?
- forget_me Forget me Deletes your saved profile and consent record immediately. Withdrawing consent is as easy as giving it. Ask: Forget me and delete my data.
- set_language Set your language Switches menus and answers to one of nine languages. It is remembered in your profile only if you have one. Ask: In German, please.
Resources and prompts
Besides tools, the server offers read-only resources that an assistant can load as background context, and ready-made prompts you can start from. Support depends on your client; the tools work everywhere.
promptquorum://categories— App categories, use cases and operating systems, with counts.promptquorum://stacks— Common app stacks by goal, with hardware floors.promptquorum://app/{slug}— One directory entry, by app slug.promptquorum://article/{cluster}/{slug}— One article, by cluster and slug.promptquorum://methodology— How to read the data: verified, unknown and estimated.
Ready-made prompts
start— Open the PromptQuorum menu (/start)recommend_local_ai_stack— Recommend a local AI stack for my hardwarecompare_local_ai_tools— Compare local AI tools for a use casefind_offline_alternative— Find a fully offline alternative to a cloud toolcheck_commercial_license— Check whether I can use an app commercially
How to read the answers
- Directory data is editorial and can be out of date. Check the official site before you install.
- "Not researched" or "unknown" means PromptQuorum has no verified figure. It does not mean the answer is no, and it does not mean it will work.
- Memory and VRAM figures for models are estimates, not benchmarks, and say nothing about speed.
- Hands-on findings keep their evidence labels (observed, measured, vendor claim, tester view). Apps we did not test are reported as untested.
- License explanations are general information, not legal advice. An app's license is separate from the license of any model you run in it.
Access and limits
Public, read-only and free, with no sign-in. A fair-use limit of 120 requests per minute per IP address applies. Usage counters are aggregate only; for the rate limit, a hashed IP address is kept for one minute and never linked to what you ask.
Live usage
Aggregate, PII-free call counters — no IP address, user agent, or other identifier is ever recorded, only how many times each tool has been called. Generated Oct 17, 2026, 3:19 AM UTC.
Total calls, all time
1,859
| Tool | Total calls | Calls today (UTC) |
|---|---|---|
| find_best_apps | 3 | 0 |
| search_apps | 119 | 0 |
| recommend_stack | 1 | 0 |
| find_local_alternative | 0 | 0 |
| get_app_alternatives | 0 | 0 |
| get_app_details | 244 | 161 |
| list_categories | 9 | 0 |
| compare_apps | 12 | 0 |
| check_hardware_compatibility | 1 | 0 |
| estimate_vram | 1 | 0 |
| explain_license | 4 | 0 |
| search_promptquorum | 490 | 6 |
| get_article | 957 | 152 |
| find_related_content | 0 | 0 |
| get_hands_on_test | 2 | 0 |
| get_latest | 1 | 0 |
| list_clusters | 11 | 0 |
| start | 4 | 0 |
| set_consent | 0 | 0 |
| my_profile | 0 | 0 |
| forget_me | 0 | 0 |
| set_language | 0 | 0 |
Raw JSON: /api/mcp-status. Directory data is editorial, may be outdated, and download links are not verified by PromptQuorum.